Home telemonitoring and remote feedback between clinic visits for asthma.

(1)

Cochrane

Database of Systematic Reviews

Home telemonitoring and remote feedback between clinic

visits for asthma (Review)

Kew KM, Cates CJ

Kew KM, Cates CJ.

Home telemonitoring and remote feedback between clinic visits for asthma. Cochrane Database of Systematic Reviews 2016, Issue 8. Art. No.: CD011714. DOI: 10.1002/14651858.CD011714.pub2.

www.cochranelibrary.com

(2)

T A B L E O F C O N T E N T S 1 HEADER . . . . 1 ABSTRACT . . . . 2 PLAIN LANGUAGE SUMMARY . . . .

4 SUMMARY OF FINDINGS FOR THE MAIN COMPARISON . . . .

7 BACKGROUND . . . . 8 OBJECTIVES . . . . 8 METHODS . . . . Figure 1. . . 10 12 RESULTS . . . . Figure 2. . . 15 18 DISCUSSION . . . . 21 AUTHORS’ CONCLUSIONS . . . . 21 ACKNOWLEDGEMENTS . . . . 21 REFERENCES . . . . 31 CHARACTERISTICS OF STUDIES . . . . 66 DATA AND ANALYSES . . . .

Analysis 1.1. Comparison 1 Home telemonitoring with feedback vs usual monitoring, Outcome 1 Exacerbations requiring oral corticosteroids (subgrouped by age). . . 67 Analysis 1.2. Comparison 1 Home telemonitoring with feedback vs usual monitoring, Outcome 2 Exacerbations requiring

ED visit (subgrouped by age). . . 68 Analysis 1.3. Comparison 1 Home telemonitoring with feedback vs usual monitoring, Outcome 3 Exacerbations requiring

hospital admission (subgrouped by age). . . 69 Analysis 1.4. Comparison 1 Home telemonitoring with feedback vs usual monitoring, Outcome 4 Asthma control

(ACQ). . . 70 Analysis 1.5. Comparison 1 Home telemonitoring with feedback vs usual monitoring, Outcome 5 Asthma control

(ACT). . . 70 Analysis 1.6. Comparison 1 Home telemonitoring with feedback vs usual monitoring, Outcome 6 ACT > 19 (well

controlled). . . 71 Analysis 1.7. Comparison 1 Home telemonitoring with feedback vs usual monitoring, Outcome 7 Asthma-related quality

of life (AQLQ). . . 72 Analysis 1.8. Comparison 1 Home telemonitoring with feedback vs usual monitoring, Outcome 8 Lung function (trough

FEV1). . . 73 Analysis 1.9. Comparison 1 Home telemonitoring with feedback vs usual monitoring, Outcome 9 Lung function (change

in PEF L/min). . . 74 Analysis 1.10. Comparison 1 Home telemonitoring with feedback vs usual monitoring, Outcome 10 Unscheduled

healthcare visits. . . 74 Analysis 2.1. Comparison 2 Type of technology subgroups, Outcome 1 Exacerbations requiring oral corticosteroids. 75 Analysis 2.2. Comparison 2 Type of technology subgroups, Outcome 2 Exacerbations requiring ED visit. . . 76 Analysis 2.3. Comparison 2 Type of technology subgroups, Outcome 3 Exacerbations requiring hospital admission. . 78 79 ADDITIONAL TABLES . . . . 84 APPENDICES . . . . 86 CONTRIBUTIONS OF AUTHORS . . . . 86 DECLARATIONS OF INTEREST . . . . 86 SOURCES OF SUPPORT . . . . 87 DIFFERENCES BETWEEN PROTOCOL AND REVIEW . . . .

i Home telemonitoring and remote feedback between clinic visits for asthma (Review)

(3)

[Intervention Review]

Home telemonitoring and remote feedback between clinic

visits for asthma

Kayleigh M Kew1_{, Christopher J Cates}1

1_{Population Health Research Institute, St George’s, University of London, London, UK}

Contact address: Kayleigh M Kew, Population Health Research Institute, St George’s, University of London, Cranmer Terrace, London, SW17 0RE, UK.kkew@sgul.ac.uk.

Editorial group: Cochrane Airways Group.

Publication status and date: New, published in Issue 8, 2016. Review content assessed as up-to-date: 24 May 2016.

Citation: Kew KM, Cates CJ. Home telemonitoring and remote feedback between clinic visits for asthma. Cochrane Database of Systematic Reviews 2016, Issue 8. Art. No.: CD011714. DOI: 10.1002/14651858.CD011714.pub2.

A B S T R A C T Background

Asthma is a chronic disease that causes reversible narrowing of the airways due to bronchoconstriction, inflammation and mucus production. Asthma continues to be associated with significant avoidable morbidity and mortality. Self management facilitated by a healthcare professional is important to keep symptoms controlled and to prevent exacerbations.

Telephone and Internet technologies can now be used by patients to measure lung function and asthma symptoms at home. Patients can then share this information electronically with their healthcare provider, who can provide feedback between clinic visits. Technology can be used in this manner to improve health outcomes and prevent the need for emergency treatment for people with asthma and other long-term health conditions.

Objectives

To assess the efficacy and safety of home telemonitoring with healthcare professional feedback between clinic visits, compared with usual care.

Search methods

We identified trials from the Cochrane Airways Review Group Specialised Register (CAGR) up to May 2016. We also searched www.clinicaltrials.gov, the World Health Organization (WHO) trials portal and reference lists of other reviews, and we contacted trial authors to ask for additional information.

Selection criteria

We included parallel randomised controlled trials (RCTs) of adults or children with asthma in which any form of technology was used to measure and share asthma monitoring data with a healthcare provider between clinic visits, compared with other monitoring or usual care. We excluded trials in which technologies were used for monitoring with no input from a doctor or nurse. We included studies reported as full-text articles, those published as abstracts only and unpublished data.

1 Home telemonitoring and remote feedback between clinic visits for asthma (Review)

(4)

Data collection and analysis

Two review authors screened the search and independently extracted risk of bias and numerical data, resolving disagreements by consensus.

We analysed dichotomous data as odds ratios (ORs) while using study participants as the unit of analysis, and continuous data as mean differences (MDs) while using random-effects models. We rated evidence for all outcomes using the GRADE (Grades of Recommendation, Assessment, Development and Evaluation Working Group) approach.

Main results

We found 18 studies including 2268 participants: 12 in adults, 5 in children and one in individuals from both age groups. Studies generally recruited people with mild to moderate persistent asthma and followed them for between three and 12 months. People in the intervention group were given one of a variety of technologies to record and share their symptoms (text messaging, Web systems or phone calls), compared with a group of people who received usual care or a control intervention.

Evidence from these studies did not show clearly whether asthma telemonitoring with feedback from a healthcare professional increases or decreases the odds of exacerbations that require a course of oral steroids (OR 0.93, 95% confidence Interval (CI) 0.60 to 1.44; 466 participants; four studies), a visit to the emergency department (OR 0.75, 95% CI 0.36 to 1.58; 1018 participants; eight studies) or a stay in hospital (OR 0.56, 95% CI 0.21 to 1.49; 1042 participants; 10 studies) compared with usual care. Our confidence was limited by imprecision in all three primary outcomes. Evidence quality ratings ranged from moderate to very low. None of the studies recorded serious or non-serious adverse events separately from asthma exacerbations.

Evidence for measures of asthma control was imprecise and inconsistent, revealing possible benefit over usual care for quality of life (MD 0.23, 95% CI 0.01 to 0.45; 796 participants; six studies; I2_{= 54%), but the effect was small and study results varied. Telemonitoring}

interventions may provide additional benefit for two measures of lung function.

Authors’ conclusions

Current evidence does not support the widespread implementation of telemonitoring with healthcare provider feedback between asthma clinic visits. Studies have not yet proven that additional telemonitoring strategies lead to better symptom control or reduced need for oral steroids over usual asthma care, nor have they ruled out unintended harms. Investigators noted small benefits for quality of life, but these are subject to risk of bias, as the studies were unblinded. Similarly, some benefits for lung function are uncertain owing to possible attrition bias.

Larger pragmatic studies in children and adults could better determine the real-world benefits of these interventions for preventing exacerbations and avoiding harms; it is difficult to generalise results from this review because benefits may be explained at least in part by the increased attention participants receive by taking part in clinical trials. Qualitative studies could inform future research by focusing on patient and provider preferences, or by identifying subgroups of patients who are more likely to attain benefit from closer monitoring, such as those who have frequent asthma attacks.

P L A I N L A N G U A G E S U M M A R Y

What are the benefits and harms of using technology to monitor people with asthma from home? Take-home message

A wide range of technologies have been developed to connect people with asthma to their healthcare professionals between routine checkups. Studies that have tested these strategies have not proved that ’telemonitoring’ leads to better symptom control or fewer attacks, and could not rule out the possibility that it may cause unintended harm by making people less likely to take action when it is needed. Telemonitoring may have small benefits for quality of life and lung function, but these results are very uncertain.

Background

Regular contact with a doctor or an asthma nurse is important to keep on top of asthma symptoms and to change inhalers if necessary. Telephone and Internet technologies are now used for lots of long-term health conditions as a way of monitoring symptoms between visits to a clinic. For asthma, lung function and other asthma symptoms can be measured at home and information sent electronically to the doctor or nurse, who can decide whether action needs to be taken before the person is due to come back to the clinic.

(5)

Review question

We wanted to find out whether home telemonitoring including feedback from a healthcare professional offers added benefits for people with asthma compared with their usual monitoring.

Study characteristics

We found 18 studies including a total of 2268 people: 12 included adults, five included children and one included individuals from both age groups. Most people included in the studies had mild to moderate persistent asthma, and studies generally lasted between three and 12 months. People in the intervention group were given one of a variety of technologies to record and share their symptoms (text messaging, Web systems or phone calls) and were compared with a group of people who received usual care, or a control group.

Main results and quality of the evidence

We could not tell whether people in the telemonitoring groups had a higher or lower chance than people in the control group of having attacks that would require a course of oral steroids, a visit to the emergency department or a hospital stay. No reports described other potential harms of home telemonitoring. Studies used lots of different types of technology, and we couldn’t tell whether some were better than others. Our confidence in the results ranged from moderate to very low, meaning that additional studies are likely to change some of these results and may influence how much we believe them.

Using technology to monitor people with asthma from home may offer benefits over usual care for overall quality of life, but the effect was small, and studies did not agree with each other. These interventions may provide benefits for lung function, but lots of people dropped out of the studies, so we couldn’t be sure.

(6)

S U M M A R Y O F F I N D I N G S F O R T H E M A I N C O M P A R I S O N [Explanation]

Home telemonitoring and feedback vs usual care for people with asthma Patient or population: people with asthm a

Setting: hom e

Intervention: hom e telem onitoring with rem ote f eedback f rom a healthcare prof essional Comparison: usual m onitoring

Outcomes Anticipated absolute effects* (95% CI) Relative effect

(95% CI)

Number of participants (studies)

Quality of the evidence (GRADE)

Comments

Risk with usual moni-toring

Risk with home tele-monitoring and feed-back

Exacerbations requir-ing oral corticosteroids

7.3-m onth f ollow-up* * 399 per 1000 382 per 1000 (285 to 489) OR 0.93 (0.60 to 1.44) 466 (4 RCTs) ⊕⊕ LOWa,b,c

2 child studies, 2 adult studies. Subgroup dif -f erences not signi-f icant (P value = 0.78) 2 child studies and 6 adult studies in ED anal-ysis agreed with the OCS analysis (OR 0.75, 95% CI 0.36 to 1.58)

Exacerbations requir-ing hospital admission

7.8-m onth f ollow-up Children (< 16 years) OR 1.38 (0.51 to 3.68) 421 (4 RCTs) ⊕⊕⊕ M ODERATEc,d

4 child studies and 6 adult studies presented separately owing to signif icant subgroup dif -f erences (P value = 0. 04)

Telem onitoring benef i-cial f or adults, but prob-ably not f or children 38 per 1000 52 per 1000 (20 to 127) Adults (17 to 65 years) OR 0.24 (0.06 to 0.94) 621 (6 RCTs) ⊕⊕⊕ M ODERATEc,d,e 83 per 1000 21 per 1000 (5 to 79) H o m e te le m o n it o ri n g a n d re m o te fe e d b a c k b e tw e e n c lin ic v is it s fo r a st h m a (R e v ie w ) C o p y ri g h t © 2 0 1 6 T h e C o c h ra n e C o lla b o ra ti o n . P u b lis h e d b y Jo h n W ile y & S o n s, L td .

(7)

Asthma control

Follow-up varied f rom 3 to 12 m onths

Asthm a control was reported in 3 dif f erent ways across 4 studies Sum m ary of results in Com -m ents colu-m n - (4 RCTs) ⊕ VERY LOWf,g,h ACQi_{(M D -0.24, 95% CI} -0.72 to 0.24) (2 adult studies); ACT ’well-controlled’ (29/ 60 vs 8/ 29) (1 adult study) (M D 0.09, 95% CI 0.92 to 1.10) (1 child study)

Serious and non- seri-ous adverse events

None of the studies explicitly reported serious or non-serious adverse events as an outcom e separate f rom asthm a exacerbation outcom es

- (0 RCTs) N/ A No studies

Asthma- related quality of life (AQLQ)i 9.6-m onth f ollow-up 1 to 7, higher = better

M ean AQLQ score was 3.58* * *

M ean AQLQ score in the intervention group was 0.23 better (0.01 better to 0.45 better) - 796 (6 RCTs) ⊕⊕ LOWf,i Lung function % predicted trough FEV1 higher = better 7.6-m onth f ollow-up

M ean predicted FEV1 was 68.4%* * *

M ean % predicted FEV1 in the intervention group was 7.21% higher (1.52 higher to 12.89 higher) - 149 (3 RCTs) ⊕⊕⊕ M ODERATEj -Unscheduled health-care visits 6.2-m onth f ollow-up 332 per 1000 329 per 1000 (155 to 565) OR 0.99 (0.37 to 2.62) 430 (3 RCTs) ⊕ VERY LOWc,k,l Very un-balanced dropout in 1 study showing dif f erent ef f ects f rom the other 2 (12% vs 57%)

* The risk in the intervention group (and its 95% conf idence interval) is based on the assum ed risk in the com parison group and the relative effect of the intervention (and its 95% CI).

* * With the exception of the asthm a control outcom e, where a range is given, f ollow-ups are given in m onths as a weighted m ean duration of studies in the analysis

* * * Risk with usual m onitoring was calculated as a weighted m ean of scores in the control groups of studies contributing to the analysis. For the AQLQ analysis, this did not include the 2 studies reporting change f rom baseline.

ACQ = Asthm a Control Questionnaire; ACT = Asthm a Control Test; AQLQ = Asthm a Quality of Lif e Questionnaire; CI = conf idence interval; FEV1= f orced expiratory volum e in 1 second; M D = m ean dif f erence; OR = odds ratio; RCT = random ised control trial; RR = risk ratio

H o m e te le m o n it o ri n g a n d re m o te fe e d b a c k b e tw e e n c lin ic v is it s fo r a st h m a (R e v ie w ) C o p y ri g h t © 2 0 1 6 T h e C o c h ra n e C o lla b o ra ti o n . P u b lis h e d b y Jo h n W ile y & S o n s, L td .

(8)

GRADE Working Group grades of evidence

High quality: We are very conf ident that the true ef f ect lies close to that of the estim ate of ef f ect

M oderate quality: We are m oderately conf ident in the ef f ect estim ate: The true ef f ect is likely to be close to the estim ate of ef f ect but m ay be substantially dif f erent Low quality: Our conf idence in the ef f ect estim ate is lim ited: The true ef f ect m ay be substantially dif f erent f rom the estim ate of ef f ect

Very low quality: We have very little conf idence in the ef f ect estim ate: The true ef f ect is likely to be substantially dif f erent f rom the estim ate of ef f ect

a_{A couple of studies carrying < 20% of the overall weight had high attrition and uncertainty with selection bias, but this was} not judged to be serious enough to downgrade (no downgrade)

b_{One study could not be included because exacerbations were used as the unit of analysis rather than people with} exacerbations, and the sm all num ber of studies in the analysis com pared with the em ergency departm ent and hospital exacerbations analyses suggests that this m ay have been recorded and not reported (-1 publication bias)

c_{Conf idence intervals include im portant benef it of either treatm ent, so it is dif f icult to interpret the result (-1 im precision)} d_{Risk of bias was conf ined m ostly to the blinding dom ains, which is unlikely to have af f ected this outcom e. Uncertainty in the} selection bias dom ains was not deem ed serious enough to downgrade (no downgrade)

e_{Heterogeneity between studies in the adult subgroup was high but not statistically signif icant, and all but one of the point} estim ates lay in the sam e direction, f avouring telem onitoring (no downgrade)

f_{Studies were generally at high risk of bias f or the blinding dom ains, which m ay have af f ected results on subjective rating} scales (-1 risk of bias)

g_{Serious inconsistency between the two studies reporting the ACQ and results across asthm a control outcom es did not give} a clear direction of ef f ect (-1 inconsistency)

h_{Im precision varied across the 3 asthm a control outcom es, but overall the ef f ect was unclear owing to dif f erences in direction,} m agnitude and conf idence intervals (-1 im precision)

j_{Child and adult studies were pooled, and im portant heterogeneity was noted (I}2_{= 54%, P value = 0.06) (-1 inconsistency)} k_{Two studies carrying m ost of the weight were judged to be at high risk of bias owing to high dropout; in particular,}_Cingi

2015had 57% dropout in the control group com pared with 12% in the intervention group (-1 risk of bias)

l_{Cingi showed an ef f ect in the opposite direction to that noted in the other two studies, which introduced im portant} heterogeneity (I2_{= 73%, P value = 0.03) (-1 inconsistency)}

i_{The m inim al clinically im portant dif f erence (M CID) f or both the Asthm a Control Questionnaire (ACQ) and the Asthm a Quality} of Lif e Questionnaire (AQLQ) is 0.5 units

H o m e te le m o n it o ri n g a n d re m o te fe e d b a c k b e tw e e n c lin ic v is it s fo r a st h m a (R e v ie w ) C o p y ri g h t © 2 0 1 6 T h e C o c h ra n e C o lla b o ra ti o n . P u b lis h e d b y Jo h n W ile y & S o n s, L td .

(9)

B A C K G R O U N D

Description of the condition

Asthma is a chronic disease of the airways that causes reversible inflammation and narrowing of the airways, along with mucus production (GINA 2014). These features commonly cause symp-toms of wheezing, breathlessness, chest tightness and cough, al-though symptoms vary between people and over time in terms of presence, frequency and severity (GINA 2014).

Despite the emergence and updating of several national and inter-national management guidelines recommending a range of cost-effective treatments based on frequency and severity of symptoms and exacerbations (e.g.BTS/SIGN 2014;GINA 2014), the dis-ease remains a significant cause of avoidable morbidity and mor-tality around the world (BTS/SIGN 2014;Global Asthma Report 2011;NRAD 2014). A national review of the 195 asthma deaths that occurred between February 2012 and January 2013 in the UK revealed that, in the year preceding their death, nearly one-third of these individuals had no record of seeing a general practitioner (GP), and nearly two-thirds had not had an asthma checkup in secondary care (NRAD 2014). The importance of self monitoring and regular checkups with a healthcare professional to monitor symptoms and encourage adherence to preventer inhalers is now well accepted (Gibson 2002;NRAD 2014), especially for those at high risk of severe asthma attacks.

Description of the intervention

Information and communication technologies have been pro-posed as a way for patients to record and share information reg-ularly about their asthma control with a healthcare professional. This method of monitoring may identify worsening asthma be-tween consultations, prompting action, such as a medication change, or an additional visit. Remote monitoring in this way is a form of ’telehealth’, otherwise referred to as ’telecare’, ’digital health’, ’mHealth’ or ’telemedicine’, which involves “the use of in-formation and communication technologies to deliver healthcare at a distance and to support patient self-management through re-mote monitoring and personalised feedback” (Mclean 2013). Communication technologies used in health care are varied, rang-ing from simple automated reminder systems for patients to take medication or attend their appointments (Gurol-Urganci 2013) to more complex health communications sent via email (Atherton 2012), telephone systems (Cash-Gibson 2012) or text messages (de Jongh 2012); however, feedback and personalised care from a healthcare professional are important components of what can be considered telehealth. Health services around the world are considering communication technologies in their various forms as a way of managing the rising number of people with long-term health conditions, to improve health outcomes and reduce the

burden on emergency and inpatient services (Steventon 2012;UK Department of Health 2012).

Governments and healthcare providers are increasingly adopting ’telehealth’ and investing in research to pin down how and for whom it could be beneficial. Programmes include an initiative in the UK encouraging wide availability of ’e-consultations’ and home ’telemonitoring’ (UK Department of Health 2013). A Tele-health Pilots Programme in Australia (Australian Government Department of Health 2016) offers widespread eHealth to chron-ically ill people and vulnerable elderly people in the Netherlands (Government of the Netherlands 2016), along with recognition of ’telehealth’ monitoring by US Medicaid insurance (Medicaid 2016). Studies have assessed the role of a range of technology-based consultations and monitoring in asthma and other health conditions, including telephone calls, email contacts, text messag-ing and video conferencmessag-ing (Laver 2013;McLean 2010;McLean 2011).

Remote monitoring of asthma with technologies might include features such as recording symptoms online or automatically trans-ferring home peak flow readings to a doctor or nurse. Regular recording and remote sharing of this information may trigger a response from a healthcare professional, who uses the information to provide personalised care. Researchers have assessed telehealth in several ways, including as an alternative for usual primary or secondary care clinic appointments (e.g.Rasmussen 2005); this was recently addressed by a related Cochrane review (Kew 2016). However, this review will consider evidence for home telemoni-toring of asthma control between visits with personalised feedback from a healthcare professional (e.g.Ryan 2012).

How the intervention might work

In the context of asthma, a condition affecting more than 300 million people worldwide (Global Asthma Report 2011), which places a significant burden on healthcare systems, telehealth may represent an unobtrusive and efficient way of maintaining contact between patients and healthcare professionals. Regular monitor-ing with communication technologies may serve to enhance self management behaviours that have known benefits for morbidity and mortality, such as keeping personalised action plans up-to-date and adhering to maintenance medications (NRAD 2014). As an alternative to methods of monitoring that do not include feed-back from a healthcare professional, telehealth may offer a more interactive method of supporting self management.

Although governments and health services have highlighted the potential for cost-savings and improved clinical outcomes of tele-health used in this way, its use to monitor patients with potentially serious or life-threatening conditions may not be without hazard. Focus groups have suggested that technology may be acceptable to patients and clinicians, but they have raised concerns that it could actually discourage self management, or might increase the 7 Home telemonitoring and remote feedback between clinic visits for asthma (Review)

(10)

likelihood of serious outcomes by instilling a false sense of security (Pinnock 2007a).

The feasibility of home telemonitoring using technology in differ-ent situations and populations may be hampered by barriers, in-cluding insufficient healthcare infrastructure and funding (Lustig 2012). However, this approach may reduce inequality in health care related to socioeconomic status and rural living by improving access to services (Jannett 2003;Lustig 2012).

Why it is important to do this review

The release of the UK National Health Service (NHS) mandate in 2013 has resulted in a push to advance the use of communication technologies for economic and clinical benefit. A recent overview of systematic reviews suggested that these benefits should not be assumed, and that people at highest risk of serious health outcomes are likely to show the biggest gains (Mclean 2013). For asthma, existing reviews have noted a large degree of variation in the way telehealth is delivered in studies and to whom and with what it is compared (Jaana 2009;McLean 2010), and have been limited for this reason in the conclusions that could be drawn. This re-view considers evidence for ongoing personalised feedback from a healthcare professional using home telemonitoring between visits, compared with monitoring without feedback. A related review has considered evidence for remote checkups as an alternative to face-to-face asthma consultations (Kew 2016).

O B J E C T I V E S

To assess the efficacy and safety of home telemonitoring with healthcare professional feedback between clinic visits, compared with usual care.

M E T H O D S

Criteria for considering studies for this review

Types of studies

We included parallel randomised controlled trials (RCTs) of any duration. We included studies reported as full-text articles, those published as abstracts only and unpublished data.

Types of participants

We included studies of adults or children with a diagnosis of asthma. We excluded studies recruiting participants with other

long-term health conditions, unless investigators presented data for people with asthma separately.

Types of interventions

We included studies comparing home telemonitoring of asthma between clinic visits, using any form of technology (e.g. telephone calls, emails, text messages, online software), with a form of mon-itoring that does not include ongoing remote professional feed-back. We included studies that compared the two types of mon-itoring on top of education or another co-intervention. We ex-cluded studies using automated telehealth interventions that did not include personalised input from a healthcare professional.

Types of outcome measures

Primary outcomes

1. Exacerbations requiring oral corticosteroids*.

2. Asthma control (measured on a validated scale, e.g. the Asthma Control Questionnaire).

3. Serious adverse events (including mortality).

Secondary outcomes

1. Asthma-related quality of life (measured on a validated scale, e.g. the Asthma Quality of Life Questionnaire (AQLQ)).

2. Unscheduled healthcare visits.

3. Lung function (trough forced expiratory volume in one second (FEV1) preferred).

4. Adverse events/side effects.

Reporting in the study of one of more of the outcomes listed here was not an inclusion criterion for the review.

*If studies reported exacerbations in a different way (e.g. requiring an emergency department (ED) visit), we analysed these separately.

Search methods for identification of studies

Electronic searches

We identified trials from the Cochrane Airways Group Specialised Register (CAGR), which is maintained by the Information Spe-cialist for the Group. This Register contains trial reports identified through systematic searches of bibliographic databases, including the Cochrane Central Register of Controlled Trials (CENTRAL), MEDLINE, EMBASE, the Cumulative Index to Nursing and Al-lied Health Literature (CINAHL), the AlAl-lied and Complemen-tary Medicine Database (AMED) and PsycINFO, and by hand-searching of respiratory journals and meeting abstracts (please see Appendix 1for further details). We searched all records in the CAGR using the search strategy presented inAppendix 2.

(11)

We also conducted a search of ClinicalTrials.gov (www.ClinicalTrials.gov) and the World Health Organization (WHO) trials portal (www.who.int/ictrp/en/). We searched all databases from their inception to May 2016, and we imposed no restriction on language of publication.

Searching other resources

We checked the reference lists of all primary studies and review articles for additional references. We searched relevant manufac-turers’ websites for trial information.

On 2 August 2016, we searched for errata or retractions from included studies published in full text on PubMed ( www.ncbi.nlm.nih.gov/pubmed).

Data collection and analysis

Selection of studies

Two review authors (KMK and CJC) independently screened ti-tles and abstracts for inclusion of all potential studies identified as a result of the search and coded them as ’retrieve’ (eligible or po-tentially eligible/unclear) or ’do not retrieve’. We retrieved the full-text study reports/publications, and two review authors (KMK and CJC) independently screened these documents, identified studies for inclusion and identified and recorded reasons for exclusion of ineligible studies. We resolved disagreements through discussion. We identified and excluded duplicates and collated multiple ports of the same study, so that each study, rather than each re-port, was the unit of interest in the review. We recorded the selec-tion process in sufficient detail to complete a PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) flow diagram (Figure 1) and aCharacteristics of excluded studiestable (Moher 2009).

(12)

Figure 1. Study flow diagram.

(13)

Data extraction and management

We used a data collection form for study characteristics and out-come data that had been piloted on at least one study in the review. One review author (KMK) extracted the following study charac-teristics from the included studies.

1. Methods: study design, total duration of study, details of any ’run-in’ period, number of study centres and locations, study setting, withdrawals and date of study.

2. Participants: N, mean age, age range, gender, severity of condition, diagnostic criteria, baseline lung function, smoking history, inclusion criteria and exclusion criteria.

3. Interventions: intervention, comparison, concomitant medications and excluded medications.

4. Outcomes: primary and secondary outcomes specified and collected and time points reported.

5. Notes: funding for trial and notable conflicts of interest of trial authors.

Two review authors (KMK and CJC) independently extracted outcome data from the included studies. We noted in the Characteristics of included studiestable if outcome data were not reported in a useable way. We resolved disagreements by consen-sus. One review author (KMK) transferred data into the Review Manager 5 (RevMan 2014) file. We double-checked that data were entered correctly by comparing data presented in the systematic review with data provided in the study reports. A second review au-thor (CJC) spot-checked study characteristics for accuracy against the trial report.

Assessment of risk of bias in included studies

Two review authors (KMK and CJC) independently assessed risk of bias for each study using the criteria outlined in the Cochrane

Handbook for Systematic Reviews of Interventions (Higgins 2011). We resolved disagreements by discussion and assessed risk of bias according to the following domains.

1. Random sequence generation. 2. Allocation concealment.

3. Blinding of participants and personnel. 4. Blinding of outcome assessment. 5. Incomplete outcome data. 6. Selective outcome reporting. 7. Other bias.

We graded each potential source of bias as high, low or unclear, and we provided a quote from the study report together with a justification for our judgement in the ’Risk of bias’ table. We sum-marised risk of bias judgements across different studies for each of the domains listed. We considered blinding separately for dif-ferent key outcomes when necessary (e.g. for unblinded outcome

assessment, risk of bias for all-cause mortality may be very differ-ent than for a patidiffer-ent-reported pain scale). When information on risk of bias relates to unpublished data or correspondence with a trialist, we noted this in the ’Risk of bias’ table.

When considering treatment effects, we took into account the risk of bias for studies that contributed to those outcomes.

Assessment of bias in conducting the systematic review

We conducted the review according to the published protocol and reported deviations from it in theDifferences between protocol and reviewsection of the systematic review.

Measures of treatment effect

We analysed dichotomous data as odds ratios (ORs), and contin-uous data as mean differences (MDs) or standardised mean differ-ences (SMDs). We entered data presented as a scale with a consis-tent direction of effect.

We undertook meta-analyses only when this was meaningful, i.e. when treatments, participants and the underlying clinical question were similar enough for pooling to make sense.

We did not include in the meta-analyses skewed data that were reported as medians and interquartile ranges, but described the results narratively instead.

When a single trial reported multiple trial arms, we included only the relevant arms. When two comparisons (e.g. drug A vs placebo, drug B vs placebo) were combined in the same meta-analysis, we halved the control group to avoid double-counting.

Unit of analysis issues

For dichotomous outcomes, we used participants, rather than events, as the unit of analysis (i.e. number of adults admitted to hospital, rather than number of admissions per adult). However, if exacerbations were reported as rate ratios, we analysed them on this basis.

Dealing with missing data

We contacted investigators or study sponsors to verify key study characteristics and to obtain missing numerical outcome data when possible (e.g. when a study was identified as abstract only). When this was not possible, and missing data were thought to introduce serious bias, we explored the impact of including such studies in the overall assessment of results by performing a sensi-tivity analysis.

(14)

Assessment of heterogeneity

We used the I² statistic to measure heterogeneity among the stud-ies in each analysis. If we identified substantial heterogeneity, we reported it and explored possible causes through prespecified sub-group analysis.

Assessment of reporting biases

When we were able to pool more than 10 studies, we created and examined a funnel plot to explore possible small-study and publication biases.

Data synthesis

We used a random-effects model for all analyses, as we expected variation in effects due to differences among study populations and interventions. We performed sensitivity analyses by using a fixed-effect model.

’Summary of findings’ table

We created a ’Summary of findings’ table by using the seven out-comes specified above. We used the five GRADE considerations (study limitations, consistency of effect, imprecision, indirectness and publication bias) to assess the quality of a body of evidence as it relates to the studies that contributed data to the meta-analyses for prespecified outcomes. We followed methods and recommen-dations described in Section 8.5 and Chapter 12 of the Cochrane

Handbook for Systematic Reviews of Interventions (Higgins 2011) while usingGRADEpro GDT 2015software. We justified all de-cisions to downgrade or upgrade the quality of studies by using footnotes, and we made comments to aid the reader’s understand-ing of the review, when necessary.

Subgroup analysis and investigation of heterogeneity

We planned the following subgroup analyses for the primary out-comes, provided at least one study was included for each subgroup.

1. Mean age (≤ 16 years, 17 to 65 years, > 65 years). 2. Type of technology (telephone calls, text messages, emails). We used the formal test for subgroup interactions provided in Review Manager 5 (RevMan 2014).

Sensitivity analysis

We planned to carry out sensitivity analyses, while excluding the following from the primary analyses.

1. Studies recruiting people with severe or life-threatening asthma.

2. Unpublished data (obtained from trial authors or from conference abstracts).

3. Studies at high risk of detection bias*.

*Inadequate selection procedures may result in unbalanced base-line characteristics between groups, which could skew the data. In light of the nature of the studies, we anticipated that all or most studies would be at high risk of performance or detection bias, so we have discussed the possible effects of lack of blinding, in particular for subjective outcomes.

R E S U L T S

Description of studies

Results of the search

We identified 619 records in the main electronic database search. We identified a total of 750 additional records through a search conducted for an older teleheathcare review with a broader scope (McLean 2010) (n = 709), clinicaltrials.gov (n = 29) and the World Health Organization (WHO) trials portal (n = 11), as well as ref-erence lists of other reviews (n = 1). Collating all searches revealed a total of 1369 records, of which 518 were duplicates. We screened the remaining 851 unique records and excluded 685 by looking at titles and abstracts alone. We reviewed full-text articles for the remaining 166 records; 123 records related to 84 studies did not meet the inclusion criteria and were excluded (with reasons), three records related to two ongoing studies (Ahmed 2011;Perry 2015) and one record is awaiting classification (Ricci 2001). This left 18 studies, with 40 associated reports, which met the inclusion criteria for this review (see trial flow inFigure 1).

Included studies

Eighteen studies, including a total of 2268 participants, met the inclusion criteria for this review (Bateman 2000; Cingi 2015; Deschildre 2012;Donald 2008;Finkelstein 2005;Guendelman 2002;Jan 2007;Kokubu 1999;Kokubu 2000;Liu 2011;Ostojic 2005; Prabhakaran 2009; Ryan 2012; van der Meer 2009; Voorend-van Bergen 2015;Willems 2008;Xu 2010;Young 2012). An overview of study, participant and intervention characteristics is given inTable 1, and more in-depth information and risk of bias details can be found in theCharacteristics of included studies table.

All included studies were parallel RCTs. The number of partici-pants in each study ranged from 16 to 288, and the median num-ber was 120. As shown inTable 1, seven studies took place in Europe (the Netherlands, Croatia, France, Turkey and the UK), five in Asia (Japan, Singapore and Taiwan), three in the USA, two in Australia and one in South Africa. Eleven studies were run from respiratory clinics in hospitals or outpatient centres (Cingi 2015;Deschildre 2012; Donald 2008;Guendelman 2002; Jan 2007;Kokubu 2000;Liu 2011;Ostojic 2005;Prabhakaran 2009; 12 Home telemonitoring and remote feedback between clinic visits for asthma (Review)

(15)

Willems 2008; Xu 2010), and four from general practitioners’ offices or medical centres (Bateman 2000;Kokubu 1999;Ryan 2012;van der Meer 2009).Young 2012was run through pharma-cies in an 11-county region in the USA, and two studies that were reported only as conference abstracts did not reveal the setting in which they took place (Finkelstein 2005;Voorend-van Bergen 2015).

Population characteristics and inclusion criteria

Twelve studies recruited adults or adults and adolescents (Bateman 2000;Cingi 2015;Donald 2008;Finkelstein 2005;Kokubu 1999; Kokubu 2000;Liu 2011;Ostojic 2005;Prabhakaran 2009;Ryan 2012; van der Meer 2009;Young 2012), five studies recruited only children (Deschildre 2012; Guendelman 2002; Jan 2007; Voorend-van Bergen 2015; Xu 2010) and one study recruited both adults and children (Willems 2008). Two of the adult studies also recruiting adolescents over 12 (Ostojic 2005;Ryan 2012) and the one study recruiting both adults and children (Willems 2008) were classified as adult studies because the mean age of participants was over 18, and data for children were not reported separately. The overall weighted mean of population ages was 31.6 years (range, seven to 52.8). The mean age of child populations was 10.4 years (range, seven to 12.1) and the mean age in adult studies was 41.6 (range, 24.7 to 52.8). The mean percentage male indicated a relatively even split of males and females (45.6% male), although the percentage male in individual studies ranged from 23.5% to 74.0%.

Deschildre 2012,Kokubu 1999,Kokubu 2000andPrabhakaran 2009listed inclusion criteria that would have led to recruitment of people with severe asthma; all required that participants had at least a course of oral steroids, a visit to the ED or admission to hospital within the previous year, and these studies excluded par-ticipants with mild or intermittent asthma or specifically required them to meet the criteria for severe asthma. Otherwise, studies gen-erally recruited people with mild to moderate persistent asthma, and common inclusion criteria included physician-diagnosed or guideline-diagnosed asthma, a recent prescription for asthma con-troller medications - usually inhaled corticosteroid (ICS) or long-acting beta agonist (LABA) + ICS - access to and competency with relevant technologies and proficiency in the given language (usually English). Common exclusion criteria were pregnancy or breastfeeding, other chronic illnesses and current smoking. Two child studies (Deschildre 2012andVoorend-van Bergen 2015) specifically recruited children with allergic asthma, andRyan 2012 required that participants score 1.5 or lower on the Asthma Con-trol Questionnaire to indicate insufficient symptom conCon-trol.

Interventions and comparisons

Six trials provided three- or four-month interventions, five tri-als gave six-month interventions and seven tritri-als tested the

in-terventions for a year (see Table 1). All monitoring interven-tions involved ways for participants or their parents to track their asthma control at home and to share this information with a healthcare professional to receive management advice between usual clinic visits. Nine studies used an Internet-based device, programme or website for participants to record and transmit a range of symptom, medication or lung function data to the health-care professional (Bateman 2000; Deschildre 2012;Finkelstein 2005;Guendelman 2002;Jan 2007;Kokubu 1999;Kokubu 2000; Voorend-van Bergen 2015; Willems 2008). Healthcare profes-sionals, often a specialist nurse, regularly reviewed the data and responded with management advice, often based on personalised asthma action plans. Six studies used a similar system of recording and response that was done primarily via short message service (SMS) or mobile phone software (Cingi 2015;Liu 2011;Ostojic 2005;Prabhakaran 2009;Ryan 2012;van der Meer 2009). Three studies involved regular calls or email contact with a nurse or phar-macist to monitor symptoms and advise on changes to medication (Donald 2008;Xu 2010;Young 2012).

Most included trials used usual care as their comparison group (Bateman 2000; Cingi 2015; Deschildre 2012; Donald 2008; Finkelstein 2005; Jan 2007; Kokubu 1999; Kokubu 2000; Prabhakaran 2009;Voorend-van Bergen 2015;Willems 2008;Xu 2010;Young 2012). In three of these studies, people in the usual care group received a minimal intervention such as an education session, a personalised asthma action plan or a peak flow meter to encourage self monitoring at home (Donald 2008;Jan 2007; Xu 2010). Five studies gave participants in the control group an asthma diary or a peak flow meter to record their symptoms at home (Guendelman 2002;Liu 2011;Ostojic 2005;Ryan 2012; van der Meer 2009), but these data were not shared with a health-care professional between usual visits.

Excluded studies

After reviewing the full texts, a total of 126 citations (86 studies) were not included in the review. It was often difficult to tell whether a study met the inclusion criteria for the review by reading the title and abstract alone, so we excluded 122 citations (83 studies) after viewing full texts. We classified three citations (two studies) as ongoing (Ahmed 2011;Perry 2015), and one citation as awaiting classification because we did not have enough details to confirm whether it met the review’s inclusion criteria (Ricci 2001). Of the 122 citations (83 studies) that were listed as excluded because they did not meet the inclusion criteria, we excluded 29 (23 studies) because closer inspection showed that inves-tigators were assessing an intervention other than home tele-monitoring, including psychological or parenting interventions (Aaron 2016;Chandler 1990;Chen 2013;Cicutto 2009;Clark 2007; Clarke 2014; Eakin 2012; Gustafson 2012; Halterman 2012; Huang 2013; Janevic 2012; Jerant 2003; Khan 2003; Kojima 2005;Lobach 2013; McCowan 2001;NCT01117805; 13 Home telemonitoring and remote feedback between clinic visits for asthma (Review)

(16)

Osman 1997;van den Berg 2002;van Gaalen 2012;van Reisen 2010; Wiecha 2007; Zachgo 2002). We excluded 25 citations (six studies) because researchers assessed the feasibility of replac-ing face-to-face reviews with reviews conducted usreplac-ing technol-ogy, which was a different question from the one we set out to answer in this review (Chan 2007; Gruffydd-Jones 2005; Hashimoto 2011; Pinnock 2003; Pinnock 2007; Rasmussen 2005). We excluded 17 citations (14 studies) because investiga-tors were assessing the use of information technologies to deliver asthma education (Barbanel 2003;Burbank 2012;De Vera 2014; Dwinger 2013;Garbutt 2010;McPherson 2006;NCT00562081; NCT00910585;NCT00964301;Pedram 2012;Peruccio 2005; Seid 2012;Shanovich 2009;Yun 2013), 19 citations (11 stud-ies) because researchers assessed automated feedback interventions that did not include ongoing input from a healthcare professional (Andersen 2007; Bender 2010; Kattan 2006; Merchant 2016; Morrison 2014; Petrie 2012; Rikkers-Mutsaerts 2012; Searing 2012; Vasbinder 2013;Vollmer 2006; Zairina 2015), nine ci-tations (seven studies) because investigators were reporting val-idation of a technology-delivered asthma questionnaire (Bender 2001;Bender 2007;Price 2007;Rand 2005;Rosenzweig 2008;

Schatz 2010;Uysal 2013) and seven citations (seven studies) be-cause the intervention was aimed purely at improving adherence rather than monitoring asthma control (Boyd 2014; Burkhart 2002; Bynum 2001; Chatkin 2006; Foster 2014; MacDonell 2015;Taitel 2014). Ten citations (nine studies) were not reports of RCTs and were recorded as using the wrong design for the re-view (Apter 2000;Araujo 2012;Claus 2004;Cruz-Correia 2007; Fonseca 2006;Friedman 1999;Lam 2011;Murphy 2001;Raat 2007), and six citations (six studies) compared a home telemoni-toring intervention with another active comparator that was not eligible for this review (Apter 2015;Baptist 2013;de Jongste 2008; NCT00149474;Schatz 2003;Sparrow 2005).

Risk of bias in included studies

Figure 2presents an overview of risk of bias in the included stud-ies, and we provide below a summary of possible bias related to each domain. We have given full details of the rationale for each judgement in each study’s risk of bias table (see theCharacteristics of included studiestables).

(17)

Figure 2. Risk of bias summary: review authors’ judgements about each risk of bias item for each included study.

(18)

Allocation

Much uncertainty surrounded the two selection bias domains, despite efforts to clarify procedures with study authors. Seven studies were at low risk of bias for random sequence generation (Cingi 2015;Ostojic 2005;Prabhakaran 2009;Ryan 2012;van der Meer 2009;Voorend-van Bergen 2015;Willems 2008), and eight were at low risk for allocation concealment (Cingi 2015; Guendelman 2002;Jan 2007;Kokubu 1999;Kokubu 2000;Ryan 2012; Voorend-van Bergen 2015;Young 2012); we considered onlyCingi 2015,Ryan 2012andVoorend-van Bergen 2015to be at low risk in both selection bias domains. We rated the remaining studies as unclear.

Blinding

It was not possible to blind participants and personnel to group allocation because of the nature of the interventions and compar-isons, and this posed the most serious risk of bias for the evidence in this review. We assessed risk of performance bias separately for objective (e.g. exacerbations) and subjective (e.g. quality of life) outcomes to better represent how this bias was likely to have af-fected our confidence in the results. We considered all studies to be at low risk of performance bias for objective outcomes and at high risk of bias for subjective outcomes.

Although theoretically outcome assessors could have been inde-pendent from the study and blinded to allocation, we did not as-sume that this was the case unless it was explicitly described in the report, or unless study authors confirmed this through per-sonal communication. Fifteen studies did not describe methods to blind outcome assessors and did not confirm that those measuring outcomes were not blinded to group allocation; we rated these as having high risk of bias (Bateman 2000;Cingi 2015;Deschildre 2012;Finkelstein 2005;Guendelman 2002;Jan 2007;Kokubu 1999;Kokubu 2000;Liu 2011;Ostojic 2005;Prabhakaran 2009; van der Meer 2009; Voorend-van Bergen 2015; Willems 2008; Xu 2010). We rated the remaining three studies as having low risk of bias (Donald 2008;Ryan 2012;Young 2012). Researchers de-scribedCingi 2015as a double-blind trial, but participants, who self rated their symptoms for the primary outcome, could have conceivably worked out which group they were in by noting what they received during the study.

Incomplete outcome data

We considered half of the included studies to be at low risk of attrition bias because attrition was low and balanced across inter-vention and control groups, or because we believed that meth-ods used to replace data for participants who did not complete the study would have adequately controlled for bias (Guendelman

2002;Jan 2007;Kokubu 2000;Ostojic 2005;Prabhakaran 2009; Ryan 2012;Voorend-van Bergen 2015;Xu 2010;Young 2012). We rated three studies as unclear because we could not tell how many people dropped out of the study. We considered six studies to be at high risk of bias, mostly because attrition was high or was much higher in one group than in another, or because analyses did not include those who dropped out before the end of the study. Willems 2008reported that up to 28% of data for particular out-comes were missing owing to errors with data transmission or poor compliance with questionnaires, andCingi 2015analysed only data for participants who completed the questionnaire at the end of the study, which represented a much smaller proportion of the control group than the intervention group (42.6% and 88.2%, respectively).

Selective reporting

We rated 14 studies as having low risk of bias because we were satisfied that all planned outcomes had been fully reported in pub-lished reports, or because study authors provided us with addi-tional information upon request. We ratedCingi 2015as low risk, although we could not meta-analyse some outcomes because of the way they were reported, and statistical methods used were appro-priate for the study data. We were unsure of the risk of reporting bias inKokubu 1999because it was available only in Japanese, and it was difficult to confirm whether all intended outcomes had been reported, even with translation. ForKokubu 2000, the translation confirmed that not all named outcomes had been reported suffi-ciently to include them in meta-analyses, so we rated this study as having high risk of bias. We rated two other studies (Bateman 2000andFinkelstein 2005) as having high risk of bias because they were available only as conference abstracts, not in peer-reviewed journals, and this meant that study authors provided very little information about the conduct of the studies or their numerical results.

Other potential sources of bias

We did not note any additional sources of bias, so we rated all studies as having low risk for ’other sources of bias’.

Effects of interventions

See:Summary of findings for the main comparison Summary of findings table 1

Primary outcomes

(19)

Exacerbations requiring oral corticosteroids

Home telemonitoring with feedback might be better or worse than usual monitoring (odds ratio (OR) 0.93, 95% confidence Interval (CI) 0.60 to 1.44; 466 participants; four studies; I2= 0%; Analysis 1.1). Over seven months, 399 people per thousand who were monitored in their usual way had an exacerbation compared with 382 per thousand if they received telemonitoring with remote feedback from a healthcare professional (95% CI 285 to 489 per 1000). We had low confidence in the estimate because only four studies could be included in the analysis (Deschildre 2012;Donald 2008;Ryan 2012;Xu 2010), indicating possible publication bias. In addition, confidence intervals were too wide to reveal whether one monitoring strategy is likely to be better than another. In addition to the four studies that were meta-analysed, Voorend-van Bergen 2015reported the total number of exacerbations rather than the number of participants having at least one exacerbation, so we could not combine their data with data from other studies. Researchers observed 10 exacerbations among the 90 participants in the Web group and 17 exacerbations in the 87 participants receiving standard care.

Eight studies reported exacerbations that required a visit to the ED (Donald 2008;Guendelman 2002;Kokubu 2000;Liu 2011; Ryan 2012;van der Meer 2009;Willems 2008;Xu 2010), which supported the main analysis that home telemonitoring and feed-back might be better or worse than control (OR 0.75, 95% CI 0.36 to 1.58; 1018 participants; eight studies; I2_{= 47%;}_Analysis 1.2). We noted important heterogeneity in the ED overall analysis and within each child and adult subgroup. We had low confidence in the effect because confidence intervals were too wide to show whether one strategy is likely to be better than another, and be-cause we noted important heterogeneity both within and across subgroups.

A look at exacerbations requiring hospital admission revealed that the overall pooled effect including child and adult studies showed uncertainty in relation to benefit or harm compared with usual monitoring (OR 0.56, 95% CI 0.21 to 1.49; 1042 participants; 10 studies; I2_{= 45%;}_{Analysis 1.3}_{). However, the test for subgroup}

differences was statistically significant, suggesting possible benefit for adults in reducing the number of exacerbations requiring hos-pital admission (OR 0.24, 95% CI 0.06 to 0.94).

Asthma control

Most studies did not use validated measures of asthma control, and the four that did could not be pooled in a single analysis (Cingi 2015;Ryan 2012;van der Meer 2009;Voorend-van Bergen 2015). Two adult studies (Ryan 2012;van der Meer 2009) used the Asthma Control Questionnaire (ACQ), for which the mini-mal clinically important difference (MCID) is 0.5. These studies showed very different effects, which made the pooled result diffi-cult to interpret (mean difference (MD) -0.24, 95% CI -0.72 to 0.24; 478 participants; two studies; I2_{= 91%).}

One child study (Voorend-van Bergen 2015) using the Asthma Control Test (ACT) as a continuous variable found no difference in scores between the two types of monitoring (MD 0.09, 95% CI -0.92 to 1.10). One adult study (Cingi 2015) reported the number of people who were classed on the ACT as ’well controlled’. The effect favoured home telemonitoring and feedback, but the CI included the possibility that the effect may be the same as that for usual monitoring (OR 2.46, 95% CI 0.94 to 6.41).

Overall we had very low confidence in asthma control outcomes owing to inconsistency both within outcomes and between them in terms of direction and magnitude of effects. In addition, impre-cision in most of the estimates made them difficult to interpret, and the nature of these subjective scales means that they may be subject to performance and detection biases associated with in-ability to blind the interventions.

Serious adverse events (including mortality)

None of the studies recorded serious adverse events separately from asthma exacerbations, and none reported whether anyone died during the study.

Subgroup analysis and investigation of heterogeneity

Mean age (< 16 years, 17 to 65 years, > 65 years)

Two child studies (Deschildre 2012;Xu 2010) and two adult stud-ies (Donald 2008;Ryan 2012) contributed to the primary out-come of requiring oral corticosteroids, and the test for subgroup differences did not indicate a difference in effects (I2= 0%, P value = 0.78). Two child studies (Guendelman 2002;Xu 2010) and six adult studies (Donald 2008;Kokubu 2000;Liu 2011;Ryan 2012; van der Meer 2009;Willems 2008) contributed to the exacerba-tion requiring ED visit analysis, and a large degree of heterogene-ity was evident within each subgroup (I2_{= 62%, P value = 0.11;}

I2_{= 53%, P value = 0.06). As above, the test for subgroup}

differ-ences for exacerbations requiring hospital admission analysis sug-gests that adults may fare better with home telemonitoring than children. It was not possible to subgroup the asthma control or serious adverse event outcomes to investigate the effects of age.

Type of technology (telephone calls, text messages, emails)

We divided studies by type of technology used, for the purpose of subgroup analysis, but we found nearly as many subgroups as stud-ies, so it was not possible to draw any conclusions about whether the type of technology influenced the effect on exacerbations re-quiring oral steroids (Analysis 2.1;Analysis 2.2;Analysis 2.3).

(20)

Sensitivity analyses

Studies recruiting people with severe or life-threatening asthma

Deschildre 2012,Kokubu 2000, andXu 2010contributed to the primary analyses that listed inclusion criteria requiring popula-tions with severe asthma.Deschildre 2012andXu 2010were the only child studies contributing to Analysis 1.1, and when they were removed, the pooled effect was very similar (OR 0.93, 95% CI 0.60 to 1.44 with; OR 0.90, 95% CI 0.54 to 1.50 without). RemovingKokubu 2000andXu 2010from the exacerbation re-quiring ED visit analysis did not change the interpretation (OR 0.75, 95% CI 0.36 to 1.58 with; OR 0.65, 95% CI 0.27 to 1.59 without). When all three studies (Deschildre 2012;Kokubu 2000; Xu 2010) were removed from the exacerbation requiring hospi-talisation outcome, the overall effect showed a similar magnitude but became more imprecise (OR 0.56, 95% CI 0.21 to 1.49, to OR 0.58, 95% CI 0.13 to 2.56). These studies did not contribute to the asthma control or serious adverse events primary analyses.

Unpublished data (obtained from trial authors or from conference abstracts)

We obtained none of the data in the primary outcome analyses from trial authors or from conference abstracts, so this sensitivity analysis was not necessary.

Studies at high risk of detection bias

We considered only three studies (Donald 2008; Ryan 2012; Young 2012) to be at low risk for detection bias, and they con-tributed only to the exacerbations outcomes, which are unlikely to have been affected by this type of bias. For these reasons, we did not conduct the sensitivity analysis.

Secondary outcomes

Asthma-related quality of life

People in the telemonitoring with feedback groups scored better on the AQLQ than those monitored in the usual way (MD 0.23, 95% CI 0.01 to 0.45; 796 participants; six studies; I2_{= 54%). The}

MCID on the scale is 0.5 units. We downgraded our confidence in the result to low because of the potential for performance and detection bias in the measure, and because we noted important variation between study results.

Kokubu 2000reported a change in the Japanese Ministry of Health and Welfare asthma quality of life score. We chose not to combine this with the AQLQ data using SMD, because this change score was obtained on a different scale, and we could not find details of properties of the measure.

Lung function

Home telemonitoring with feedback showed an overall benefit on lung function compared with usual monitoring, measured as percentage predicted pre-bronchodilator forced expiratory volume in one second (FEV1) (MD 7.21, 95% CI 1.52 to 12.89; 149

participants; three studies; I2_{= 0%) and change in peak expiratory}

flow (PEF) (MD 13.20, 95% CI 0.58 to 25.82; 66 participants; one study).

van der Meer 2009reported FEV1as litres change from baseline,

which could not be pooled with results of the other studies. This study showed a 240 mL mean increase from baseline in the home telemonitoring group (SD 810 mL) and a 10 mL mean decrease in the control group (SD 752 mL). Additionally,Voorend-van Bergen 2015reported several lung function parameters as z-scores in the paper; we could not use the absolute final scores, as we observed a significant baseline imbalance between groups.

Unscheduled healthcare visits

Variation between study results made the pooled effect difficult to interpret (OR 0.99, 95% CI 0.37 to 2.62; 430 participants; three studies; I2_{= 73%), but telemonitoring did not lead to a clear}

increase or decrease in the number of people making unscheduled healthcare visits. We had very low confidence in the result owing to imprecision of the estimate, attrition bias and important het-erogeneity.

Deschildre 2012reported data as the mean number of unsched-uled visits per participant, which could not be pooled with di-chotomous data. The study showed a slightly higher rate of un-scheduled visits in the home telemonitoring group (mean 5.24, SD 3.62) compared with the usual care group (mean 4.43, SD 4.13).

Adverse events/side effects

As with serious adverse events, none of the studies explicitly re-ported adverse events as an outcome separate from asthma-related adverse outcomes.

D I S C U S S I O N

Summary of main results

Home telemonitoring with feedback might be better or worse than usual monitoring for exacerbations requiring a course of oral steroids (odds ratio (OR) 0.93, 95% confidence interval (CI) 0.60 to 1.44; 466 participants; four studies), a visit to the emergency department (OR 0.75, 95% CI 0.36 to 1.58; 1018 participants; eight studies) or a hospital stay (OR 0.56, 95% CI 0.21 to 1.49; 18 Home telemonitoring and remote feedback between clinic visits for asthma (Review)

(21)

1042 participants; 10 studies). Our confidence in the results was reduced owing to wide confidence intervals, which meant that we could not rule out important benefits or harms of the intervention. Evidence for measures of asthma control was patchy and did not show a consistent direction of effect, with most studies not using validated measures that could be pooled. We noted imprecision in the estimates, and the nature of these subjective scales means they may be subject to performance and detection bias associated with inability to blind the interventions.

None of the studies recorded serious or non-serious adverse events separately from asthma exacerbations, and none reported whether anyone died during the studies; this is a limitation of the evidence. Researchers have been concerned that this type of increased mon-itoring can lead to a false sense of security, actually increasing ad-verse events and the need for emergency care, which does not seem to be the case. However, the benefits of home telemonitoring are modest at best, given the resources and infrastructure required to implement them.

With the exception of hospital admissions, adult and child stud-ies showed similar findings, and too few studstud-ies with too much variation in the interventions used prevented any meaningful con-clusions about which types of technology may offer the greatest benefit.

Within the secondary outcomes, people in the telemonitoring groups scored better on the Asthma Quality of Life Questionnaire (AQLQ) than those monitored in the usual way (mean difference (MD) 0.23, 95% CI 0.01 to 0.45; 796 participants; six studies; I

2_{= 54%), but study results showed important variation, and even}

the upper CIs did not reach what is considered to be a meaningful difference on the scale (0.5 units). Some benefit of home telemon-itoring on lung function was apparent.

Overall completeness and applicability of evidence

The field of telehealthcare is rapidly growing, and national health systems are pushing to bring home telemonitoring interventions into widespread practice; this explains the number of studies that met our inclusion criteria. We deliberately refined the question in such a way that we would exclude studies of automated monitor-ing based on algorithms, monitormonitor-ing as part of broad telehealth interventions that included all manner of education and adherence modules and sharing of symptom data to inform asthma man-agement. That said, some of the included study interventions did involve components that may have confounded the comparison we set out to measure (e.g. an asthma education page within a monitoring website, clinician decision support to interpret mon-itoring data). We focused on telemonmon-itoring interventions that required clinician input to make the evidence easier to apply to real clinical situations, but the number of excluded studies illus-trates the breadth and complexity of this growing field of health care, along with the evidence that has not been considered in this

review. Moreover, communication technologies for interventions with another primary focus such as education, compliance or in-haler technique have not been considered, and this splitting may cause difficulty for decision makers in a field that is as varied and dynamic as the technologies themselves (Rada 2015).

Some of the adult studies recruited people with severe or uncon-trolled asthma (Deschildre 2012;Kokubu 1999;Kokubu 2000; Prabhakaran 2009;Ryan 2012), but most did not specify, and we found a mix both within and across analyses. This has impli-cations for services that seek to implement monitoring strategies such as these, because we cannot be sure whether people who have regular exacerbations would derive the greatest benefit from extra monitoring, or in fact whether they are more likely to be harmed by the weakened responsibility patients feel in terms of their own health. A related issue is the small number of studies reporting what we considered to be the most important outcomes (i.e. those prespecified in our review protocol), in particular, lack of explicit reporting of adverse events, which may be related to the frequency of these events expected in different populations. Other sources of variation such as the nature of the control group (e.g. provision of an action plan inRyan 2012) and the frequency of planned and actual feedback provided by healthcare professionals also make the results difficult to apply to practice. We did not seek to explore uptake, acceptability, equity of access and persistence in use, all of which are important determinants of how an intervention may be applied to a real-life setting.

Quality of the evidence

Our confidence in the evidence based on GRADE (Grades of Rec-ommendation, Assessment, Development and Evaluation Work-ing Group) quality ratWork-ings ranged from moderate to very low, with none of the analyses thought to be of high quality.

When considering the effects of risk of bias on evidence quality, we noted that most of the high risk of bias judgements involved the blinding domains, which reduced our confidence only for subjec-tive outcomes (asthma control and quality of life). We also noted some issues with high or unbalanced dropout in some studies, al-though we considered this to have affected only the lung function and unscheduled healthcare visits analyses, for which high-risk studies carried a lot of weight. For one of the primary outcomes -exacerbations requiring oral steroids - a couple of the contributing studies had high attrition and uncertainty with selection bias, but they carried less than a fifth of the overall weight, so we did not judge this as serious enough to warrant a downgrade.

Inconsistency between study results was a problem in several of the analyses (asthma control, exacerbations requiring emergency department (ED) visits and hospital admissions, quality of life and unscheduled healthcare visits) and could not be explained in most cases by planned subgroup analyses for age or type of technology. This may reflect the variation in any number of factors related to the studies, such as baseline characteristics of recruited popula-19 Home telemonitoring and remote feedback between clinic visits for asthma (Review)